Model comparison
Codestral vs GPT-5.6 Terra
GPT-5.6 Terra is the stronger model overall, scoring 59.1 to 30.1 on the Noometry Index. Codestral costs 10× less per token, which makes it the better buy when GPT-5.6 Terra's lead doesn't matter for your workload.
Last verified . 2 shared benchmarks.
Mistral AISummary
- They share 2 benchmarks with published results for both. Codestral scores higher in 0 categories and GPT-5.6 Terra in 2 categories; 2 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.6 Terra leads 60.6 to 19.1.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 32.5% for Codestral and 51.3% for GPT-5.6 Terra.
- Codestral is cheaper at $0.30 / $0.90 per million input/output tokens, against $2 / $12 for GPT-5.6 Terra.
- GPT-5.6 Terra accepts more context: 1.05M tokens versus 256K.
- Codestral has downloadable open weights; the other is API-only.
Side by side
| Codestral | GPT-5.6 Terra | |
|---|---|---|
| Provider | Mistral AI | OpenAI |
| Noometry Index | 30.1 | 59.1 |
| Released | 2024-05-29 | 2026-07-09 |
| Weights | Open | Proprietary |
| Context window | 256K | 1.05M |
| Max output | 4K | 128K |
| Input $ / M tokens | $0.30 | $2 |
| Output $ / M tokens | $0.90 | $12 |
| Results tracked | 7 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Terra leads
Codestral: 26.9 (#316), GPT-5.6 Terra: 57.1 (#16)
| Benchmark | Codestral | GPT-5.6 Terra |
|---|---|---|
| ALE-Bench | 137.78 | 1,951 |
| DeepSWE | — | 69.6% |
| FrontierCode | — | 41.3% |
| Aider Polyglot | 11.1% | — |
| CursorBench | — | 41.3% |
| LMArena WebDev | — | 1520 |
| SciCode | — | 55% |
| WeirdML | — | 78.3% |
| BigCodeBench Instruct | 41.8% | — |
| LMArena Coding | — | 1484 |
| BigCodeBench Complete | 52.5% | — |
| HumanEval+ | 73.8% | — |
| MBPP+ | 61.9% | — |
Agentic & Tool Use Not comparable
Codestral: —, GPT-5.6 Terra: 40.2 (#23)
| Benchmark | Codestral | GPT-5.6 Terra |
|---|---|---|
| APEX-Agents | — | 58.2% |
| BALROG | — | 53.2% |
| GDP.pdf | — | 24.7% |
| Vending-Bench 2 | — | 7,343 |
Reasoning GPT-5.6 Terra leads
Codestral: 19.1 (#244), GPT-5.6 Terra: 60.6 (#19)
| Benchmark | Codestral | GPT-5.6 Terra |
|---|---|---|
| Kagi LLM Benchmark | 32.5% | 51.3% |
| ARC-AGI-2 | — | 83.9% |
| SimpleBench | — | 48.9% |
| NYT Connections (extended) | — | 78.4% |
| ARC-AGI-1 | — | 96.5% |
| CritPt | — | 30% |
| Chess Puzzles | — | 54% |
| LMArena Hard Prompts | — | 1467 |
| Mystery Game Puzzles | — | 35% |
| DTBench | — | 93.3% |
| LMCA | — | 55% |
| Surface Evolver Bench | — | 83.8% |
| Epoch Capabilities Index | — | 159.31 |
Math Not comparable
Codestral: —, GPT-5.6 Terra: 81.5 (#10)
| Benchmark | Codestral | GPT-5.6 Terra |
|---|---|---|
| FrontierMath (Tiers 1-3) | — | 86% |
| FrontierMath Tier 4 | — | 70.7% |
| OTIS Mock AIME 2024-2025 | — | 99.7% |
| ProofBench | — | 74% |
| LMArena Math | — | 1463 |
Knowledge Not comparable
Codestral: —, GPT-5.6 Terra: 61.3 (#27)
| Benchmark | Codestral | GPT-5.6 Terra |
|---|---|---|
| GPQA Diamond | — | 93.3% |
| SimpleQA Verified | — | 43.2% |
| LMArena Expert | — | 1492 |
Multimodal Not comparable
Codestral: —, GPT-5.6 Terra: 47.5 (#10)
| Benchmark | Codestral | GPT-5.6 Terra |
|---|---|---|
| LMArena Vision | — | 1271 |
| Blueprint-Bench 2 | — | 30.8% |
| Furniture Assembly | — | 54.2% |
| LMArena Document | — | 1472 |
Multilingual Not comparable
Codestral: —, GPT-5.6 Terra: 54.5 (#42)
| Benchmark | Codestral | GPT-5.6 Terra |
|---|---|---|
| LMArena Non-English | — | 1439 |
| LMArena Chinese | — | 1514 |
| LMArena French | — | 1476 |
| LMArena German | — | 1462 |
| LMArena Japanese | — | 1446 |
| LMArena Korean | — | 1420 |
| LMArena Russian | — | 1450 |
| LMArena Spanish | — | 1455 |
Instruction Following Not comparable
Codestral: —, GPT-5.6 Terra: 76.6 (#38)
| Benchmark | Codestral | GPT-5.6 Terra |
|---|---|---|
| LMArena Instruction Following | — | 1453 |
Long Context Not comparable
Codestral: —, GPT-5.6 Terra: 44.4 (#64)
| Benchmark | Codestral | GPT-5.6 Terra |
|---|---|---|
| LMArena Longer Query | — | 1452 |
Writing & Preference Not comparable
Codestral: —, GPT-5.6 Terra: 70.4 (#21)
| Benchmark | Codestral | GPT-5.6 Terra |
|---|---|---|
| LMArena Text | — | 1446 |
| LMArena Creative Writing | — | 1411 |
| EQ-Bench Creative Writing | — | 1855 |
| EQ-Bench 4 | — | 1234 |
| LMArena Multi-Turn | — | 1451 |
Frequently asked questions
Is Codestral better than GPT-5.6 Terra?
GPT-5.6 Terra is the stronger model overall, scoring 59.1 to 30.1 on the Noometry Index. Codestral costs 10× less per token, which makes it the better buy when GPT-5.6 Terra's lead doesn't matter for your workload.
Which is cheaper, Codestral or GPT-5.6 Terra?
Codestral is cheaper. It lists at $0.30 per million input tokens and $0.90 per million output tokens; GPT-5.6 Terra lists at $2 and $12.
Is Codestral or GPT-5.6 Terra better for coding?
GPT-5.6 Terra scores higher on coding benchmarks: 57.1 versus 26.9 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Terra does, with 1.05M tokens against 256K.
How many benchmarks do Codestral and GPT-5.6 Terra share?
2 benchmarks have published results for both models. Codestral has 7 scored results on Noometry and GPT-5.6 Terra has 52.